Developing A Framework to Support Human Evaluation of Bias in Generated Free Response Text
This work addresses bias evaluation for LLM deployments in real-world contexts, though it appears incremental as it builds on existing human evaluation approaches with automation enhancements.
The authors tackled the challenge of evaluating bias in LLM-generated free text responses by developing a semi-automated framework that integrates human insights, resulting in an operational definition of bias and a methodology for classification beyond multiple-choice benchmarks.
LLM evaluation is challenging even the case of base models. In real world deployments, evaluation is further complicated by the interplay of task specific prompts and experiential context. At scale, bias evaluation is often based on short context, fixed choice benchmarks that can be rapidly evaluated, however, these can lose validity when the LLMs' deployed context differs. Large scale human evaluation is often seen as too intractable and costly. Here we present our journey towards developing a semi-automated bias evaluation framework for free text responses that has human insights at its core. We discuss how we developed an operational definition of bias that helped us automate our pipeline and a methodology for classifying bias beyond multiple choice. We additionally comment on how human evaluation helped us uncover problematic templates in a bias benchmark.